Dataset with 200 patients and severe class imbalance (2% event rate), representing extremely rare outcomes where PPV/NPV are critical metrics.
Format
A data frame with 200 rows and 3 variables:
- patient_id
Character: Patient identifier (PT001-PT200)
- rare_outcome
Factor: "Event" or "No_Event" (2%/98% prevalence)
- predictor
Numeric: Predictor value (mean: 90 for event, 50 for no event)
Details
Extreme class imbalance (2% events) with good predictor discrimination. Tests handling of severely imbalanced data and emphasis on predictive values over sensitivity/specificity.
Examples
data(psychopdaROC_imbalanced)
psychopdaROC(data = psychopdaROC_imbalanced, dependentVars = "predictor",
classVar = "rare_outcome", positiveClass = "Event",
refVar = "predictor")
#>
#> ADVANCED ROC ANALYSIS
#>
#>
#>
#>
#> Procedure Notes
#>
#>
#>
#> The ROC analysis has been completed using the following
#> specifications:
#>
#>
#>
#> Measure Variable(s): predictor
#>
#> Class Variable: rare_outcome
#>
#> Positive Class: Event
#>
#>
#>
#> Method: maximize_metric
#>
#> All Observed Cutpoints: FALSE
#>
#> Metric: youden
#>
#> Direction (relative to cutpoint): >=
#>
#> Tie Breakers: mean
#>
#> Metric Tolerance: 1e-06
#>
#>
#>
#> <hr />
#>
#> <div style='padding: 10px; background-color: #f8f9fa; border: 1px
#> solid #dee2e6; border-radius: 4px; margin-bottom: 15px;'>
#>
#> Analysis Status
#>
#> Seed: 123Positive Class: Event (Prevalence: 2%)Analysis Mode:
#> Basic<div style='background-color: #fff3cd; color: #856404; padding:
#> 10px; border-radius: 4px; margin-top: 10px;'>Warnings:Class imbalance
#> detected (Prevalence: 2.0%). Consider using Precision-Recall curves.
#>
#> ROC Analysis Summary
#> ────────────────────────────────────────────────────────────────────────
#> Variable AUC 95% CI Lower 95% CI Upper p-value
#> ────────────────────────────────────────────────────────────────────────
#> predictor 0.9553571 0.9235097 0.9872046 < .0000001
#> ────────────────────────────────────────────────────────────────────────
#> Note. Reading of the test values: <b>HIGHER values were taken to
#> indicate Event</b> (Classification Direction = ">="). If that is
#> the wrong way round for this marker, every sensitivity,
#> specificity, cutpoint and AUC below is reversed — switch
#> Classification Direction to "<=" and the AUC becomes 1 minus the
#> value shown.
#> Note. AUC 95% confidence intervals computed using the DeLong
#> method.
#>
#>
#> Clinical Interpretation
#> ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
#> Test Performance Level Clinical Recommendation Detailed Interpretation
#> ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
#> predictor Excellent Suitable for clinical use with appropriate cutpoint The test 'predictor' has an AUC of 0.955 indicating excellent discriminatory ability. This test can reliably distinguish between diseased and healthy patients.
#> ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
#>
#>
#> OPTIMAL CUTPOINTS AND PERFORMANCE
#>
#> no title
#> ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────
#> Cutpoint Sensitivity Specificity PPV NPV Youden's J AUC Metric Score
#> ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────
#> 76.9122686 100.00000 93.36735 23.52941 100.00000 0.9336735 0.9553571 0.9336735
#> ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────
#>
#>
#> Area Under the ROC Curve
#> ────────────────────────────────────────────────────────────────────────
#> Variable AUC 95% CI Lower 95% CI Upper p-value
#> ────────────────────────────────────────────────────────────────────────
#> predictor 0.9553571 0.9235097 0.9872046 < .0000001
#> ────────────────────────────────────────────────────────────────────────
#> Note. Reading of the test values: <b>HIGHER values were taken to
#> indicate Event</b> (Classification Direction = ">="). If that is
#> the wrong way round for this marker, every sensitivity,
#> specificity, cutpoint and AUC below is reversed — switch
#> Classification Direction to "<=" and the AUC becomes 1 minus the
#> value shown.
#> Note. AUC 95% confidence intervals computed using the DeLong
#> method.
#>
